未验证 提交 b8cf8ca7 编写于 作者: W WangXi 提交者: GitHub

fill_constant_batch_size_like op support fp16 (#39907)

上级 f77019a0
...@@ -18,6 +18,8 @@ limitations under the License. */ ...@@ -18,6 +18,8 @@ limitations under the License. */
namespace ops = paddle::operators; namespace ops = paddle::operators;
REGISTER_OP_CUDA_KERNEL( REGISTER_OP_CUDA_KERNEL(
fill_constant_batch_size_like, fill_constant_batch_size_like,
ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CUDADeviceContext,
paddle::platform::float16>,
ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CUDADeviceContext, ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CUDADeviceContext,
float>, float>,
ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CUDADeviceContext, ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CUDADeviceContext,
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册